Advanced digital methods for condition assessment of transformers and reactors in combination with machine learning

Inhalt:
Modern transformers or variable shunt reactors are equipped with a vacuum type on-load tap-changer, where these type of OLTCs can perform 300.000 operations before maintenance is required. Therefore, it makes sense to determine the evolution of mechanical condition of the tap-changer by a monitoring system over these very long operating periods before inspection. Vibro Acoustic Measurement (VAM) is a method that allows the assessment of the on-load tap-changer switching sequence during its operation. In the process of on-load tap-changer operation, the acoustic signal reflects mechanical and electrical events during the switching process. While, the switching sequence cannot be evaluated directly, stable acoustic timing of the switching sequence is one of the preconditions for correct and reliable tap-changer operation. The core idea of the evaluation method is the calculation of the envelope of the VAM signal, its labelling and analysis of extracted features. The acoustic events, features on the envelope curve, are not all identical to the electrical events, e.g. the opening or closing of the electrical contacts of the diverter switch, but result from the mechanical processes in the diverter switch and also from those that occur before and after the electrical switching. Some of the acoustic events are usually closely related to the electrical events. Each type of OLTC has its individual combination of usable events in the VAM signal. Because of mechanical tolerances of the OLTCs the parameter of the switching sequences, especially the internal switching times may be slightly different for different exemplars of the same asset type. That's why every OLTC has its own individual switching pattern. The influence of environmental conditions, like temperature, should also be considered to improve the interpretability of the data obtained. This paper introduces a new statistical method for analysis, visualization and interpretation of vibro-acoustic measurements of on-load tap-changer in service in combination with simple data models usable for long-term VAM monitoring. The results of a four-year field test will be presented, together with the evaluation procedures used.